The science-specific home learning environment of elementary school children – how are science experiences and science talk associated with the children’s science achievement?

ABSTRACT

Background

The home learning environment (HLE) is central to children’s development. Activities in the HLE are assumed to mediate the effects of distal factors such as family socio-economic status (SES) and parental beliefs and attitudes on children’s development. However, little is known about the science-specific HLE and the relationship between different parent-child science activities and elementary school children’s science learning.

Purpose

This study aims to investigate a) how different science activities in the HLE relate to children’s science achievement and b) how they mediate the associations of family SES and parents’ beliefs and attitudes on children’s science achievement.

Sample

Data from the German subsample of the Trends in Mathematics and Science Study (TIMSS) 2019 was analyzed, including 3,437 4th-grade children (M_age = 10.39 years, 50% boys).

Design and methods

The study has a cross-sectional design. We used structure equation modeling (SEM) to examine associations and test for indirect effects.

Results

We identified two different patterns of science activities in the HLE: science experiences (e.g. reading science books, experimenting) and science talk. SEM revealed that science talk was associated with children’s science achievement, whereas science experiences were not. Further, science talk mediated associations of family SES and parental beliefs and attitudes on the children’s science achievement.

Conclusions

Our results might indicate that discursive elements in science activities, such as parent-child science talk, are crucial for children’s science learning, whereas a mere provision of science experiences may not adequately foster the children. We suggest that future research should focus on analyzing the quality of parent-child interactions in different science activities to understand possible learning processes in these activities better. Overall, although effect sizes were small, our results emphasize the crucial role of the HLE for elementary school children’s science learning.

KEYWORDS:

• Home learning environment
• elementary/primary
• informal education
• family

Introduction

Science learning is often examined within school education. However, students spend a lot of time outside school; thus, learning in the family context may also contribute to their knowledge and attitudes about science (Lin and Schunn 2016; Otani 2020; Sammons et al. 2015). This may be especially true for elementary school students, who are often more involved in science learning within the home learning environment (HLE) than older students (Toth et al. 2020). However, little is known about how elementary school children’s HLE contributes to their science achievement.

The HLE is often conceptualized as a combination of distal and proximal factors influencing children’s learning (Grolig 2020; Kluczniok et al. 2013; Niklas and Schneider 2017). Distal factors include 1) structural family characteristics such as the socio-economic status (SES) and 2) parents’ beliefs and attitudes. These factors are assumed to influence children’s learning not directly but indirectly through home learning activities involving parents and children (Bradley and Corwyn 2002; Hoover-Dempsey and Sandler 2005; Kluczniok et al. 2013). In turn, home learning activities act as the central and proximal factor that directly influences children’s learning and mediates the effects of distal factors (Barger et al. 2019; Hill and Tyson 2009; Junge et al. 2021; Kluczniok et al. 2013; Melhuish et al. 2008; Otani 2020).

Previous work has mainly concentrated on the relationship between the HLE and children’s learning in literacy and numeracy (e.g. Anders et al. 2012; Kluczniok et al. 2013; Segers, Kleemans, and Verhoeven 2015; Skwarchuk, Sowinski, and LeFevre 2014), with limited focus on science. First studies in the domain of science suggest links between children’s science achievement and family SES (Blums et al. 2017; Zhang et al. 2019) or parents’ science beliefs and attitudes (Perera 2014). However, as the focus primarily lies on single aspects of the HLE, we lack a comprehensive understanding of the science-specific HLE and its contributions to children’s science learning.

Furthermore, the specific types of home learning activities and their effect on children’s learning remain less understood in science. Parents might foster their children in various science activities inside and outside the home, such as using science media like magazines, TV shows or internet resources together (Hightower et al. 2022; Lin and Schunn 2016), conducting small experiments, exploring nature and observing animals and plants, or visiting museums or zoos and talking about the exhibits (Crowley et al. 2001; Lin and Schunn 2016). However, if and how these home learning activities contribute to children’s science learning and if there are differential effects as known from literacy and numeracy (LeFevre et al. 2009; Manolitsis, Georgiou, and Tziraki 2013; Skwarchuk, Sowinski, and LeFevre 2014), remains poorly understood. Further, it is uncertain whether insights from other domains can be translated to science, as there may be notable differences between the domains. For example, parents might find it easier to identify beneficial and age-appropriate learning activities in mathematics and language, such as reading, spelling, writing, or counting and calculating, than in science, where activities are often less general and more topic-specific. Moreover, parents may often lack the ideas or confidence to participate in science activities with their children (Silander et al. 2018; Sonnenschein, Gursoy, and Stites 2022), or consider science to be less important than other subjects (Andre et al. 1999; Saçkes et al. 2011; Sonnenschein, Gursoy, and Stites 2022), possibly suggesting different relationships between science activities and children’s science learning as compared to other domains.

Thus, using data from the German subsample of the Trends in Mathematics and Science Study (TIMSS; Kasper et al. 2023) in 2019, this study aims to analyze the science-specific HLE of elementary school children. It focuses on the relationships between different science activities in the HLE and elementary school children’s science achievement. Further, building upon prior conceptualizations of the HLE, we seek to test the mediating role of the science activities between family’s structural characteristics and parents’ science-specific beliefs and attitudes on the children’s science achievement.

Background

Theoretical framework of the home learning environment

We frame the HLE within the bioecological theory of Bronfenbrenner and Morris (2006). This theory posits that children’s development is shaped by various interconnected systems, such as the family and school, which exert proximal influences on the child’s development and, in turn, are influenced by more distal factors. Drawing on the capital theory of Bourdieu (1986), family structural characteristics like the family’s financial, cultural, or social resources – often operationalized through family SES or parental education – serve as a distal influence on children’s learning in the HLE (Bradley and Corwyn 2002; Sirin 2005). While a lack of these resources might impact children’s development (Bourdieu 1986; Bradley and Corwyn 2002; Niklas and Schneider 2017), their influence is less significant than the proximal features of the HLE, such as joint parent-child activities (Melhuish et al. 2008; Niklas and Schneider 2017). Instead, the lack of financial resources, lower education, and limited access to cultural possessions may diminish the frequency and quality of home learning activities, which in turn affects the children’s development (Bradley and Corwyn 2002; Zady and Portes 2001; Zhang et al. 2019).

In conceptualizations of the HLE, parental beliefs and attitudes are considered as another distal factor that influences children’s development (Kluczniok et al. 2013; Skwarchuk, Sowinski, and LeFevre 2014). The theory of planned behavior provides insights into how parents’ beliefs and attitudes influence children’s learning (Ajzen 1991): Parents’ intentions to perform an action are predicted by their beliefs and attitudes towards a subject, subjective normative beliefs (i.e. the social desirability to perform the action), and their control beliefs, which are closely linked to their perceived self-efficacy to succeed in that action (Bandura 1977). Accordingly, parental beliefs and attitudes, such as their expectations for their child’s education or their interest in a specific domain, directly influence parents’ behavior in activities with their children (Ajzen 1991), which, in turn, mediate the effects of parental beliefs and attitudes on children’s development (Hoover-Dempsey and Sandler 2005; Junge et al. 2021; Kluczniok et al. 2013; Skwarchuk, Sowinski, and LeFevre 2014).

Therefore, home learning activities represent a key and proximal factor for children’s learning within the HLE. In line with socio-cultural perspectives, children learn during these activities by interacting with knowledgeable others, such as their parents (Vygotsky 1987). Parents may scaffold their children’s learning and provide support tailored to the child’s individual level to enhance their knowledge (Wood, Bruner, and Ross 1976), for example, by engaging in discursive elements with their children through open-ended questions, hints and assistance, or explanations, which are especially promotive for their learning (cf. Crowley et al. 2001; Leech et al. 2020; Tenenbaum and Leaper 2003).

Empirical research on the home learning environment

Structural characteristics

There is broad literature focusing on family SES and its association with children’s achievement in different domains like numeracy, literacy, and science, showing that SES achievement gaps already exist early in age (Kelly et al. 2011; Sirin 2005; Zhang et al. 2019) and persist into elementary school and beyond (Blums et al. 2017; Morgan et al. 2016; Saçkes et al. 2011). Evidence from the science domain suggests that the SES of families of elementary and preschool children is related to the frequency of science activities in the HLE (He et al. 2023; Junge et al. 2021). Thereby, families with higher SES tend to attach greater importance to science than other subjects (Saçkes 2014). Moreover, preschool and elementary school children from families with higher SES achieve substantially higher in science and mathematics than children from families with lower SES (Mullis et al. 2020; Zhang et al. 2019), which is partially mediated by the science activities (Junge et al. 2021). Further, children’s home language was found to be associated with children’s science achievement (Mullis et al. 2020). Children speaking a second language engage less frequently in science activities with their parents, which in turn affects their science knowledge (Junge et al. 2021; see also Kluczniok et al. 2013).

Parental beliefs and attitudes

In the literature, various parental beliefs and attitudes were shown to be associated with children’s development. A meta-analysis indicated that parental expectations, beliefs and attitudes towards the importance of education are positively associated with children’s achievement in various domains (Hill and Tyson 2009). For the science domain, Perera (2014) showed that positive parental attitudes are associated with secondary school children’s science achievement, using data from 15 countries participating in PISA 2006. Likewise, parents’ interest in science is associated with preschool children’s science achievement and, further, is mediated by the frequency of science activities at home (Junge et al. 2021). However, there is evidence that parents participate less in their children’s science learning in comparison to other domains like mathematics or literacy, although the children enjoyed science activities the most (Saçkes 2014; Sonnenschein, Gursoy, and Stites 2022). Parents receive science as less important, lack confidence or interest in science maybe because of unpleasant experiences in their own science education, and, in turn, initiate less science activities with their elementary school children (Andre et al. 1999; Kaya and Lundeen 2010; Saçkes 2014; Sonnenschein, Gursoy, and Stites 2022). Moreover, results indicate gender-specific differences. Parents tend to offer boys more science activities and explanations than girls (Alexander, Johnson, and Kelley 2012; Andre et al. 1999; Crowley et al. 2001) and prefer science activities over activities in other domains rather for boys than girls (Saçkes 2014). These tendencies were also shown for older children in 4^th to 6^th grade, with parents believing that science is more interesting for boys than for girls and fathers using more cognitive demanding speech for their sons than their daughters (Andre et al. 1999; Tenenbaum and Leaper 2003). There are, however, mixed findings for gender differences in elementary and secondary school children’s science achievement (Mullis et al. 2020; OECD 2016) and science aspirations, interest or self-efficacy (Andre et al. 1999; DeWitt, Archer, and Osborne 2014; Silver and Rushton 2008; Tenenbaum and Leaper 2003) showing no clear picture for gender differences within the science domain.

Home learning activities

Home learning activities are considered the proximal factor for children’s development. For elementary school children, activities like taking the children to libraries or sports or making learning materials accessible positively contribute to the children’s learning next to school influences (Barger et al. 2019; Hill and Tyson 2009; Sammons et al. 2015). However, two meta-analyses showed that solely parental assistance in their children’s homework is negatively associated with the children’s achievement (Barger et al. 2019; Hill and Tyson 2009). One explanation may be that parents compensate for their children’s learning deficits with additional homework assistance.

Domain-specific activities are directly associated with children’s respective domain-specific achievement (Anders et al. 2012; Dearing et al. 2012; Junge et al. 2021; Manolitsis, Georgiou, and Tziraki 2013; Skwarchuk, Sowinski, and LeFevre 2014). Further evidence suggests that different home learning activities within one domain may have differential effects on the children’s domain-specific achievement. In the literacy and numeracy domains, for instance, formal activities (e.g. teaching the children the alphabet/numbers) are distinguished from informal activities (e.g. shared book reading/playing board games), that each contributes differentially to the preschool children’s learning within these domains (LeFevre et al. 2009; Manolitsis, Georgiou, and Tziraki 2013; Skwarchuk, Sowinski, and LeFevre 2014).

However, it is less clear whether the concept of formal/informal activities can be applied to the science domain, and whether different types of activities have different effects on children’s learning.

Science activities

In science, recent research distinguishes between activities in various informal science learning environments, like everyday (e.g. gardening, jointly watching science TV shows), outside-home (e.g. visiting zoo/museum), and school-led activities (e.g. school trips) (DeWitt and Archer 2017; He et al. 2023; Lin and Schunn 2016). For preschool, elementary and middle-school school children, there is first evidence that everyday and outside-home activities contribute to children’s science learning (Eberbach and Crowley 2017; Junge et al. 2021; Kaderavek et al. 2020; Suter 2014). In everyday activities, for instance, parents use many extra-textual utterances and science talk when they read science books with their 4–7-years old children, e.g. they give explanations beyond the phenomena within that book and refer to the children’s daily life, which positively impacts children’s recognition of simple science concepts (Leech et al. 2020; Miller-Goldwater et al. 2023; Shirefley and Leaper 2021; Shirefley et al. 2020). Furthermore, Vandermaas-Peeler, Mischka, and Sands (2019) and Vandermaas-Peeler et al. (2018) showed that parents and children engage in simple science activities like cooking and baking, playing games or going out into nature. Thereby, they interact by making observations, asking questions, making predictions, or comparing and classifying. In addition, parents were shown to support their preschool and elementary school children, scaffold their learning, and provide explanations (Siegel et al. 2007; Tenenbaum and Leaper 2003; Tenenbaum et al. 2005) that presumably are beneficial for the children’s science learning also with regard to further learning at school (Kaderavek et al. 2020; Tenenbaum et al. 2005). Parents use similar support in outside-home activities (Crowley et al. 2001; Siegel et al. 2007), which is beneficial for children’s science learning (Eberbach and Crowley 2017; Suter 2014). However, unlike in literacy and numeracy, the differential contributions of various activities to children’s science learning are less systematically investigated.

The present study

The main goal of this study was to investigate the relationship between science-specific HLE and elementary school children’s science achievement. Further, we aimed to explore a) whether there are different types of science activities similar to the (in)formal activities in literacy and numeracy and b) if these are differentially associated with the children’s science achievement. Therefore, we first conducted an exploratory factor analysis (EFA) of everyday science activities. For each of these possible types of activities revealed by factor analysis, we seek to answer the following research questions: How are the activities related to the children’s science achievement? How are structural characteristics and parental beliefs and attitudes related to the activities? Do the activities mediate the effects of the distal factors on the children’s science achievement?

Method

Sample

Data was derived from the German subsample of the Trends in Mathematics and Science Study (TIMSS) in 2019 (Kasper et al. 2023). TIMSS is an international study in which quadrennially the achievement of 4^th- and 8^th-grade students in mathematics and science is assessed. Further data is obtained from student, parent, teacher, and school questionnaires (Mullis et al. 2020). However, in Germany, only the 4^th-grade elementary school students are assessed within TIMSS. In TIMSS 2019, N = 3437 4^th-graders aged 10.39 years (SD = 0.51) took part in the study, of whom 50.50% were boys. Measured by language spoken at home (i.e. children who not solely spoke German at home), 38.67% of the children had a migration background. The children came from 203 different schools and 211 different classes, respectively.

Measures

Science achievement

Children’s science achievement was derived from an assessment across the three science content domains: Earth Science, Life Science, and Physical Science (Mullis et al. 2020). Students were exposed to only a part of the question pool used in TIMSS 2019. According to planned missing data design and based on Item Response Theory, for each student and for each content domain, five plausible values were obtained using latent regression and multiple imputation techniques (Foy et al. 2017). Each of these plausible values thereby represents a score for the complete science assessment.

Science activities

In a questionnaire, parents reported on the frequency of joint everyday science activities. On a four-point Likert scale (1: never or almost never, 2: once or twice a month, 3: once or twice a week, 4: every or almost every day) parents rated five statements (originally in German language) about (a) shared reading about science, (b) talking about everyday phenomena, (c) talking about science school topics, (d) experimenting and (e) jointly watching science TV shows (Beese et al. 2022). The detailed items are given in Appendix 1. For reliability indices, we calculated Revelle’s ω, which is used when the criteria of unidimensionality as well as τequivalence of a scale are not met (as will be shown later). As recommended by Cho (2016), we will provide both ω total (ω_T, i.e. the proportion of variance explained by the general and the minor factors relative to the total variance) and ω hierarchical (ω_H, i.e. the proportion of variance only explained by a general factor underlying the whole scale and where the minor factor variances are partialized out) with ω_T = .83 and ω_H = .64 showing high reliability.

Structural characteristics

Information about family characteristics was collected in the parents’ questionnaire. We used the Highest International Socio-Economic Index of Occupational Status (HISEI) as an indicator of family SES (Ganzeboom, de Graaf, and Treiman 1992).

Parents’ science-specific beliefs and attitudes

Parents were surveyed about their science-specific beliefs and attitudes, i.e. about their view on the importance of science for society as well as the personal relevance of science (Beese et al. 2022). In a questionnaire, they rated five statements on a four-point Likert scale (1: totally disagree, 2: rather disagree, 3: rather agree, 4: totally agree). The items are given in Appendix 2. Reliability of the scale was high with ω_T = .81 and ω_H = .62.

Covariates

We controlled for the effects of children’s age, gender, cognitive abilities, and home language. Children’s cognitive abilities were assessed by the subscale series of the standardized Culture Fair Test (CFT) 20-R (Weiß 2006). Further, as an indicator of German language abilities, the language spoken at home was surveyed in the student’s questionnaire, i.e. whether the children always speak German at home (coded as 4), almost always speak German (coded as 3), whether they only sometimes speak German (coded as 2) or never (coded as 1).

Preliminary analysis

In a preliminary data analysis, EFA revealed two factors with Eigenvalues > 1 underlying the science activities scale (see Table 1 for more details). Note that in this preliminary analysis, communality h of item (b) was high and uniqueness u was low, showing a large amount of variance explained in (b) by both factors. Further, item (c) loaded considerably on both factor 1 and factor 2. Therefore, to consolidate our findings from EFA, we further ran a confirmatory factor analysis (CFA): We compared three models, a comparative baseline model with all five items loading on one latent factor and two further models with the respective item patterns revealed by EFA loading on two separate factors (see Table 2). Factor loading estimations on the first variables were each constrained to the value 1 by default. Since in Model 2 only two items (b) and (c) loaded on one factor, this model was estimated with both factor loadings τ-equalized to avoid underidentification and parameter estimation bias through Heywood-cases (Chen et al. 2001; Little, Lindenberger, and Nesselroade 1999). Considering the almost identical loadings of both items (b) and (c) in the baseline model, τ-equalizing the loadings was highly reasonable. See Table 2 for further details.

Table 1. Results of EFA for the science activities scale.

	Loadings λ
Variable	Factor 1 (λ1)	Factor 2 (λ2)	h	u
(a) Shared Reading about Science	0.54	0.26	0.37	0.6336
(b) Talking about Everyday Phenomena	0.30	0.95	1.00	0.0047
(c) Talking about Science School Topics	0.45	0.56	0.51	0.4859
(d) Experimenting	0.63	0.25	0.46	0.5359
(e) Jointly Watching Science TV Shows	0.58	0.19	0.38	0.6238

h = communality; u = uniqueness. Considerable factor loadings are in bold type.

Table 2. Results of CFA for the science activities scale. Standardized loadings within each model are shown.

	Baseline Model	Model 1	Model 2
Variable	λB	λ1	λ2
(a) Shared Reading about Science	0.551	0.622
(b) Talking about Everyday Phenomena	0.793		0.775
(c) Talking about Science School Topics	0.788		0.859
(d) Experimenting	0.588	0.662
(e) Jointly Watching Science TV Shows	0.512	0.614
AIC	37672.34	24301.07	14380.50
BIC	37733.76	24337.93	14398.93

Our results indicate improper τ-equivalence of factor loadings and poor fit statistics in the baseline model (RMSEA = .109, SRMR = .051). As Model 1 and Model 2 were saturated, no RMSEA or SRMR for assessing data fit was calculated. However, comparative fit indices (AIC and BIC) were substantially smaller for model 1 and 2, whereby lower indices represent more preferable models. Model 1 and model 2 clearly outperform the baseline model regarding τ-equivalence and comparative fit, which led us to consequently reject the baseline model. Thus, in further calculations, we used two separate measurement models for the science activities. We will refer to items (a) shared reading about science, (d) experimenting and (e) watching science TV shows as science experiences and to items (b) talking about everyday phenomena and (c) talking about science school topics as science talk. Both factors were moderately correlated (r = .56, p < 0.001).

Data analysis

Analytical process

To test our hypotheses for both science experiences and science talk, structure equation modelling (SEM) was conducted separately using the R-package lavaan (Rosseel et al. 2022). To answer our research questions, science achievement scores were regressed on science experiences and science talk, respectively, parents’ science-specific beliefs and attitudes (all as latent variables), and HISEI, whilst children’s age, home language and cognitive abilities served as control variables (all as manifest variables). We did not control for gender differences in SEM because we identified no associations between gender and any dependent variables (see next section). All prerequisites for regressions were met (linearity of relationships, normality and independence of residuals, homoscedasticity). To test mediation, science experiences or science talk were regressed on parental beliefs and attitudes and HISEI, resulting in the SEM depicted in Figure 1. To investigate inferential statistics about the indirect effects, we used a Monte Carlo approach as described by Preacher and Selig (2012). Bootstrapping or the commonly used delta method could yield biased estimates when the amount of missing data is large and is therefore not recommended, especially when imputing data (Biesanz, Falk, and Savalei 2010). Inferential statistics of indirect effects were computed using a script adapted for nested imputation from the MonteCarloCI function of the semTools package in R (Jorgensen et al. 2021).

Figure 1. Proposed model to identify direct and indirect effects between the manifest and latent variables. For each pattern of science activities identified by factor analysis, separate SEMs were calculated. Covariates are omitted in this figure.

Missing data

For handling of missings, we multiply imputed data under a missing at random (MAR) assumption (Rubin 2003). As the original TIMSS 2019 dataset contained five sets of plausible values on which further imputations relied on, we calculated 50 multiple imputations for each set. Hence, we yielded five nests with 50 complete datasets each, on which SEMs were conducted. The parameter estimates and fit indices of SEMs were then pooled according to the rules for nested multiply imputed datasets proposed by Rubin (2003). Nested multiple imputation and pooling of data was performed in R using the package miceadds (Robitzsch, Grund, and Henke 2021).

Results

Preliminary descriptive results

Subsequently, descriptive statistics and correlation coefficients between variables are presented in Table 3 in detail. Results were pooled over 250 datasets within five nests. Manifest bivariate correlations indicated that science talk was significantly associated with all other variables except gender. In contrast, science experiences were only associated with parents’ science-specific beliefs and attitudes. Further, we identified medium significant correlations between science achievement and background variables like HISEI, language, CFT, and age. Thereby, children’s age was negatively associated with most of the other variables such as achievement and cognitive abilities. This may be due to older children starting school later because of disadvantages. It is noteworthy that we did not find any significant associations between children’s gender and science achievement, science experiences or science talk.

Table 3. Descriptive statistics and bivariate correlations r between variables.

							Manifest Bivariate Correlations r
	Variable	M	SD	Min	Max	Ncmplt	1	2	3	4	5	6	7	8	9
1	ScAchievement	517.48	76.93	220.04	736.57	3437	1
2	ScExperiences	2.24	0.65	1	4	2160	−.03	1
3	Science Talk	2.88	0.73	1	4	2165	.12***	.56***	1
4	ScBeliefsAttit	3.15	0.53	1	4	2129	.11***	.15***	.17***	1
5	HISEI	52.21	20.07	14.21	88.96	2054	.38***	−.03	.10***	.09***	1
6	Language	3.37	0.87	1	4	2940	.30***	−.04	.06**	.01	.17***	1
7	CFT	100.58	3.19	87.38	108.39	3406	.44***	−.03	.05*	.07**	.21***	.12***	1
8	Age (years)	10.39	0.51	8.83	13.33	3432	−.24***	.01	−.06**	−.05*	−.19***	−.12***	−.14***	1
9	Gender (in %)	50.50				3433	.03	−.02	.01	.03	.01	.02	−.09***	.07***	1

***p < .001; **p < .01; *p < .05. N_cmplt = Number of complete cases. ScAchievement = Science Achievement, ScExperiences = Science Experiences, ScBeliefsAttit = Parents’ Science-Specific Beliefs and Attitudes, HISEI = Highest International Socio-Economic Index of Occupational Status, CFT = Culture Fair Intelligence Test. Gender: 1 = boys.

Associations between distal factors, science activities and children’s science achievement

Results from SEMs indicate that science talk was significantly associated with children’s science achievement, whereas science experiences were not. However, the effect sizes were small. We found moderate relations between family SES (HISEI), language spoken at home, cognitive abilities and children’s science achievement in both models, whereas children’s age was negatively related to their science achievement. We moreover identified small associations between parental beliefs and attitudes and children’s science achievement (β ≈ .07). However, the associations between parents’ science-specific beliefs and attitudes and both science experiences and science talk were substantially larger in both models (β ≈ .19). Further, SES was significantly and directly related to science talk, but not science experiences, although this relationship was relatively small. Overall, fit indices for both models were good with RMSEA < .05, SRMR < .05, CFI > .95, and TLI > .95 (Hu and Bentler 1999). Results are shown in Table 4, with pooled estimates from 250 nested imputed datasets.

Table 4. Results of structure equation modeling for both proposed models.

	Model 1			Model 2
Regressions	β	b	(SE)	β	b	(SE)
Science Achievement ~
HISEI	.253***	.012	(.001)	.251***	.012	(.001)
ScBeliefsAttit	.070**	.128	(.041)	.057**	.106	(.042)
Science Experiences	−.014	−.025	(.047)
Science Talk				.053*	.078	(.034)
Age	−.120***	−.223	(.036)	−.118***	−.220	(.035)
Language	.209***	.228	(.019)	.207***	.226	(.019)
CFT	.352***	.105	(.005)	.352***	.105	(.005)
Science Experiences ~
HISEI	−.050	−.001	(.001)
ScBeliefsAttit	.193***	.197	(.032)
Science talk ~
HISEI				.084***	.003	(.001)
ScBeliefsAttit				.189***	.237	(.036)
R^2			.355			.357
RMSEA			.029			.031
SRMR			.042			.043
TLI			.984			.976
CFI			.987			.989

***p < .001; **p < .01; *p < .05. R² represents the amount of explained variance in science achievement. For the measurement model of the latent variable science talk, loadings of both variables were τ-equalized. HISEI = Highest International Socio-Economic Index of Occupational Status, ScBeliefsAttit = Parents’ Science-Specific Beliefs and Attitudes, CFT = Culture Fair Intelligence Test.

Indirect effects of distal factors on children’s science achievement via science activities

We estimated indirect effects of distal factors on children’s science achievement via science activities only for science talk since we did not find any association between science experiences and science achievement. We found small, albeit significant indirect effects from HISEI on science achievement via science talk with β = .005 (corresponding CIs of CI_lower < .001 and CI_upper = .011). Likewise, parents’ science-specific beliefs and attitudes were indirectly associated with children’s science achievement via science talk (β = .010, CI_lower = .001 and CI_upper = .021). The results were the pooled statistics from 250 imputed datasets.

Discussion

This study examined the association between everyday science activities in the HLE and elementary school children’s science achievement. Further, it explored relations between family SES, parental beliefs and attitudes, and science activities. Using data from TIMSS 2019, we identified two different types of science activities: science experiences (e.g. shared reading about science, joint watching science TV shows) and science talk (talking about everyday phenomena, talking about science school topics). Analysis via SEM revealed significant relationships between science talk and children’s science achievement. For science experiences, this relationship was not significant. These results were partly in line with previous research on the relationships between the frequency of learning activities and preschool and elementary school children’s achievement in the domains of literacy and numeracy (e.g. Anders et al. 2012; Dearing et al. 2012; Kluczniok et al. 2013; Manolitsis, Georgiou, and Tziraki 2013) and science (Junge et al. 2021)

Our results suggest that the discursive elements during science activities with a knowledgeable other, such as the frequency of parent-child science talk, are crucial for children’s science learning. This is in line with socio-cultural theories (Vygotsky 1987) and previous research from both school (Leach and Scott 2002; Soysal and Yilmaz-Tuzun 2021) and family contexts (Eberbach and Crowley 2017; Leech et al. 2020; Miller-Goldwater et al. 2023) that emphasize the importance of high-quality teacher/parent-child discourse during science activities as a foundational principle for fostering children’s science learning. High-quality teacher/parent-child discourse encompasses, for example, asking open-ended questions, providing explanations, or connecting to prior experiences (Haden 2010; Miller-Goldwater et al. 2023).

In addition, studies from both school (Engle and Conant 2002) and family contexts (Gutwill and Allen 2010; Kaderavek et al. 2020; Miller-Goldwater et al. 2023) indicate that high-quality science experiences have the potential to facilitate teacher/parent-child discourse. However, simply exposing children to science experiences without engaging in discussions may not effectively promote their science learning. Therefore, science experiences need to be complemented by high-quality science talk to effectively foster children’s science learning (cf. Leach and Scott 2002).

This might explain the missing relationships between science experiences and children’s science achievement. Parents might merely provide science experiences without engaging in high-quality discourse, which is crucial in order to foster children’s learning during science activities (Eberbach and Crowley 2017; Leech et al. 2020; Miller-Goldwater et al. 2023). However, more research is needed on parent-child discourse during different science activities to better understand the quality of science experiences for the children’s learning. Moreover, the quality of media and materials parents and children interact with during science experiences may be inappropriate to promote science learning. For instance, Miller-Goldwater et al. (2023) showed that many science books for preschool children are of low quality with respect to learning. Similar findings have been shown for learning apps in science and other domains (Meyer et al. 2021). This could also be true for other media and materials, like science TV shows or experimental kits families might use during science experiences. However, further research is needed to investigate the quality of science learning media and materials like books, apps, and experimental kits and their contributions to children’s science achievement. In addition, the topics of science experiences in the HLE may differ from curricular school topics that are assessed in TIMSS 2019. For instance, Hoover-Dempsey and Sandler (2005) pointed out that some home learning activities may not be directly associated with the children’s achievement in school because the topics may not overlap. However, they may be associated with children’s interest or self-efficacy in science, which in turn relates to science achievement (see also Otani 2020). This could especially be true for science experiences, whereas science talk is, at least partly, more apposite to school topics.

Overall, this study contributes to the existing literature by indicating that the frequency of parent-child science talk relates to the children’s science achievement and that, in addition to school influences, the science-specific HLE contributes to elementary children’s science achievement. Although effect sizes between science activities and science achievement were smaller than in studies with preschool children (Junge et al. 2021), a direct comparison is difficult due to different measures. However, this might indicate that the HLE is more critical for younger than older children, for whom learning opportunities at school are particularly important (Sammons et al. 2015).

We found positive associations between family SES (measured by the HISEI) and science talk, indicating that parents with higher SES and potentially higher science knowledge might be more able to support and engage more often in high-quality discourses with their children (Bradley and Corwyn 2002; Zhang et al. 2019). This result is consistent with previous findings in the domain of science, which indicate positive linkages between family SES and the HLE for preschool and elementary school children (Blums et al. 2017; DeWitt and Archer 2017; He et al. 2023; Junge et al. 2021; Morgan et al. 2016; Saçkes et al. 2011). However, science experiences were not associated with family SES. This is surprising, as materials for science experiences, such as science books or experimental kits, might be more affordable for high-SES parents. One possible explanation might be that family resources do not play a significant role in certain science experiences, such as watching science TV shows, as operationalized in this study.

We found positive relationships between science-specific parental beliefs and attitudes and science talk and science experiences, respectively. Parents with more positive beliefs and attitudes engage more often in science talk and science experiences with their elementary school children. We add to previous research showing that parents when they value science more, have a higher interest in science, or higher self-efficacy, more often conduct science activities with their children (Alexander, Johnson, and Kelley 2012; Junge et al. 2021; Perera 2014; Sonnenschein, Gursoy, and Stites 2022), which is also in line with research in other domains (e.g. Hoover-Dempsey and Sandler 2005; Kluczniok et al. 2013; Segers, Kleemans, and Verhoeven 2015).

Further, we found that science talk mediated the effect of distal factors on children’s science achievement: Higher-SES parents engage more often, and potentially in higher-quality science talk with their children (Zady and Portes 2001), which in turn is positively associated with the children’s science achievement. Likewise, if parents have more positive science-specific beliefs and attitudes, the more science talk they engage in, which again relates to the children’s science achievement. These findings are consistent with previous research in the domains of literacy and numeracy (e.g. Dearing et al. 2012; Kluczniok et al. 2013).

In contrast to other findings (Crowley et al. 2001; Franse, van Schijndel, and Raijmakers 2020; Tenenbaum and Leaper 2003), we found no gender-specific differences in science activities in the HLE. However, our study used a survey design instead of observational data, so the results are not directly comparable.

Limitations

A number of limitations should be considered. First, we only considered a few activities that were assumed to be associated with children’s science achievement. Further science activities might also influence the children’s learning, like gardening, jointly using science apps, visiting zoos or museums (DeWitt and Archer 2017; He et al. 2023; Lin and Schunn 2016), and should be considered in future research. Moreover, this study could not analyze how the parents interacted with their children during these activities. Future studies must also account for the quality of parent-child interactions next to their mere frequency. This involves focusing on various aspects of interaction quality such as parents’ verbal and nonverbal behaviors of questioning, explaining, connecting, and socio-emotional support. As this study separately accounted for science talk and science experiences, this could also help to better understand the important role of parent-child discourses in the everyday science experiences they encounter. Another limitation is that while TIMSS affords a large and representative set of student achievement data, it is restricted by the cross-sectional design. Therefore, our study shows associations but no causal mechanisms. The reported relationships could also work in the opposite direction, such as higher achieving children demanding more science talk, which results in more science discourses, which in turn may also influence parents’ science-specific beliefs and attitudes. Additionally, it is unclear to what extent the topics of the reported science activities match the topics of the curriculum-based TIMSS assessments. Some of the science activities at home, such as watching TV shows or reading books, may address other science phenomena than those implemented at school and may, therefore, be less overlapping with the school assessments. Furthermore, our study is based on self-reports from the parents. This limits the data quality through respondents’ understanding (e.g. of a science book or experiment) and social desirability. A further limitation is the high amount of missing data in the TIMSS dataset. However, we considered missings using nested multiple imputations under a MAR assumption. To consolidate our findings and to account for possible other missing mechanisms, we conducted our analyses also with raw data showing that multiple imputation yielded very robust results. Lastly, a limitation might be that we did not control for school influences on the children’s science achievement. However, our main focus was on the influences of HLE on children’s development.

Conclusion and implications

Our findings show differential associations between science activities in the HLE and children’s science achievement. In contrast to science experiences, only science talk was related to science achievement and also mediated distal factors like family SES and parental beliefs and attributes on children’s science achievement. Thus, our work contributes to previous findings from other domains and age groups (He et al. 2023; Hightower et al. 2022; Junge et al. 2021; Kluczniok et al. 2013). It indicates that besides formal schooling, the HLE also plays a role in domain-specific (e.g. science) learning. Further work is needed to develop a broader understanding of science HLE, such as the effect of different activities and the importance of the quality of parent-child interactions during these activities for children’s science learning. On a higher level, our results suggest that interventions for parents that address beliefs and attitudes, such as the relevance of science, might be a way to promote more science activities at home (Saçkes 2014; Sonnenschein, Gursoy, and Stites 2022).

Ethics statement

All aspects of ethics were met within the settings of TIMSS.

Acknowledgments

The data were made available by the Research Data Centre at the Institute for Educational Quality Improvement (FDZ at IQB).

Disclosure statement

No potential conflict of interest was reported by the authors.

Supplemental material

Supplemental data for this article can be accessed at https://doi.org/10.1080/02635143.2024.2375512